To the contrary, it perfectly fits the dictionary definition: Discrimination on the basis of gender.

It happens to be a form of discrimination which is (generally) considered to be acceptable since it is aiming to ameliorate a disadvantaged group; but it is discrimination nonetheless. The relevant question isn't "is this sexist", but rather "is this sexist in a way that society does not condone".

The difference between this and actual sexism is that sexism discourages and systematically excludes people, typically girls/women, based on gender. Programs that target girls do nothing at all to discourage boys, because no one is saying, "These kinds of activities, careers, etc are only for girls now."

Generally speaking, when any disadvantaged group gets together to discuss issues and take action on issues affecting them, they are not being racist, sexist, etc.

Is the new statistical model: "Correlation not equal to causation unless it agrees with my politics"?

Most people don't, however, and interpret the word in its more traditional sense, which is "discriminatory on the basis of sex or gender". And in that sense, yeah, it is very much sexist. Arguably justifiable on the "lesser evil" basis, in that it tries to affect the outcomes in way that negates other (unjustified) sexism elsewhere in the system. But nevertheless, from this perspective, the fact that you have to discriminate is bad, and the ideal state of affairs is where it's not necessary (i.e. there's no sexism elsewhere to mitigate).

But the evidence of sexism is the existence of the imbalance. This whole thing is circular.